The San Diego OC DISCUSSION Thread

[Gorod]

Omg , 3300MHz thats insane ! Not many FX55 will do that Why did i bought this Venice with only 3150MHz max. overclock ? Dang i need to sell it and try a SanDiego insted

[kryptobs2000]

Wonder whats causing this big decrease in memory bandwidth. Your timings could be different than saftlad's.

[brandinb]

no the sandiego core has a drop in memory bandwidth compared to old winchester and claw cores

maybe it can be fixed in a bios upgrade
________
Ritalin Rehab Advice

[Dumo]

I wonder too...Hows fx55BN will compare to 55-130nm..

[ben805]

The San Diego need less vcore to run the same speed as the Venice, but the SD runs almost 20 degree hotter....around 60'C under load! I ran the PCmark04 to compare to his Venice and found almost no gain, cpumark99 is slightly higher that's about it, at this point I don't know if it really worth the extra money for gain that I can barely see over the Venice, though I have a single stage phase change on the way soon, will have to wait and see how well it does under extreme temperature, I hope this SD has slightly higher ceiling...


My 3700 SD:



Venice:

[Mrki]

phase change, cpu temp -13c, look at the pic

the temp diode does not read correctly

[xgman]

I wonder too...Hows fx55BN will compare to 55-130nm..

Got one coming in later this week. But then again my 130 55 does 3 gighz on water so I'm not expecting miracles.

[Soulburner]

This is going to become a problem, if a mod is going to move posts out of the OC thread and into the discussion thread, they need to edit the post and include a link to the post being referred to. This thread is just going to end up being a "dump" for all posts that don't belong in the OC thread with none of them really correlating with each other as a discussion...

[uwackme]

368 vs 332 for CPUmark result sure looks like a 10% improvement for 1M cache vs 512K cache....in this particular bench, certainly worth noting.

Also, 26m 54sec vs 26m 09sec is a nice gain for simply large cache.

[ben805]

368 vs 332 for CPUmark result sure looks like a 10% improvement for 1M cache vs 512K cache....in this particular bench, certainly worth noting.

Also, 26m 54sec vs 26m 09sec is a nice gain for simply large cache.

cpumark99 is the only benchmark that shows more than 5% gain, otherwise 3DMark, PCmark, Sandra, Winrar, and any other benchmarks didn't show much. look at the version of my SuperPI....it's the SSE3 patched version so the gain is expected, I'll run the unpatched version to show you the jaw dropping gain I got

[cpulloverclock]

cpumark99 is the only benchmark that shows more than 5% gain, otherwise 3DMark, PCmark, Sandra, Winrar, and any other benchmarks didn't show much. look at the version of my SuperPI....it's the SSE3 patched version so the gain is expected, I'll run the unpatched version to show you the jaw dropping gain I got

with my 3000+ I get 26.9s at 3046MHz unpatched version and my 4000+ has 26.9s at 2992-5MHz but I believe the FX-55 is better than the SD for SPI

-200/300MB/s of bandwidth is too bad

improvement of 1.5-2%

[ben805]

with my 3000+ I get 26.9s at 3046MHz unpatched version and my 4000+ has 26.9s at 2992-5MHz but I believe the FX-55 is better than the SD for SPI

-200/300MB/s of bandwidth is too bad

improvement of 1.5-2%

I also noticed that as well, the FX-55 also yield a noticeable higher memory bandwidth so maybe the memory controller on those FX are superior to the SD or Venice? my SD on memtest also shows couple hundreds less than the Venice and Winnie so I don't know what the deal is....maybe it's just a software or bios bugs, still toying with the idea of getting a used FX-55 to run on PS but I'll hang on to the SD for now, and hopefully it's just a software issues that can be patch with future updates. I have the Mushkin Redline XP4000 on the way so maybe it will be a different story to pair them up together.
If I had to do it all over again and if I were to stick with air or water cool, I'll definitely get the Venice, the SD is very toasty....underload is around 59~60'C, the Venice running at the same speed and higher vcore only load at 42~45'C, that's quite a different, perhaps removing the IHS should help but since I have the phase change on the way, I better keep it on for now. I'll load up the Windows 2003 Server and see how it does..

[joe2004]

Agree. I have now SD 3700 and this is not better than Venice 3000 while it is $200 more. Venice overclocks better on air and the 1 MB gain on L2 cache is too little advantage, as far as I am concerned. The difference is minimal. I can get SD to 2.75 with 1.5V and this is where it gets hot already. Venice can take 1.7V and I can get it to 2.85 GHz.
Frankly, SD is a waste of money, in my view.

[webbyboy]

Agree. I have now SD 3700 and this is not better than Venice 3000 while it is $200 more. Venice overclocks better on air and the 1 MB gain on L2 cache is nowhere to be seen, as far as I am concerned. The difference is minimal. I can get SD to 2.75 with 1.5V and this is where it gets hot already. Venice can take 1.7V and I can get it to 2.85 GHz.
Frankly, SD is a waste of money, in my view.

I wanted to get a SD too but seeing the results made me think of venice more. And i think the gain of 1MB L2 is very minimal.

[joe2004]

I think if you have a phase change cooling then this is a worthy chip. It is just simply too hot to overclock high on air.

[ben805]

Agree. I have now SD 3700 and this is not better than Venice 3000 while it is $200 more. Venice overclocks better on air and the 1 MB gain on L2 cache is too little advantage, as far as I am concerned. The difference is minimal. I can get SD to 2.75 with 1.5V and this is where it gets hot already. Venice can take 1.7V and I can get it to 2.85 GHz.
Frankly, SD is a waste of money, in my view.

Yup, if you're going to air/water cool...looks like the Venice is probably a better choice since it runs MUCH cooler, probably better chances to overclock higher as well. IMHO the 1MB L2 cache is overrated and if you have a tight budget and trying to get the most out of your money, then I don't think 3700 SD worth the extra $144 over the 3200 Venice, and of course...this is just my personal opinion.
But as Joe said, phase change and the like should combat the high heat that these SD puts out..

[ben805]

I wanted to get a SD too but seeing the results made me think of venice more. And i think the gain of 1MB L2 is very minimal.

minimal is an overstatement, give me sometime and I'll give you a side by side comparison of these two chips under PCMark04 at the same clock speed, this is my preference benchy because it gives me a better picture of an overall performance such as audio/video encoding/decoding, file compression, encryption, virus scan and stuff like that.

[brandinb]

Yup, if you're going to air/water cool...looks like the Venice is probably a better choice since it runs MUCH cooler, probably better chances to overclock higher as well. IMHO the 1MB L2 cache is overrated and if you have a tight budget and trying to get the most out of your money, then I don't think 3700 SD worth the extra $144 over the 3200 Venice, and of course...this is just my personal opinion.
But as Joe said, phase change and the like should combat the high heat that these SD puts out..

i was thinking the same thing while testing my sandiego then i poped the heat spreader and my load temps went from 56c on water to 42c with prime95 and thats with more v-core and 100mhz higher overclock!!!!!

my heatspreader had bubble gum under it it was a bad sight. i dont think they use shin etsu anymore mine had gray buble gum stuff. i lost 18c by removing it and can now run pi at 2930 when before it maxed at 2800 so......

do it!!
________
Volcano vaporizer

[eva2000]

I bought OPB's 4000+ Sandy CABHE 0515 SPMW. Anyone experience this ? OPB managed 2850mhz at 1.45v on air using X-90C but i can't get anywhere near that on water ?

superpi 1m stable only with 12x cpu multiplier at:

2700mhz at 1.52v
2760mhz at 1.57v
2808mhz at 1.65-1.67v

try <1.5v and BSOD on windows boot

on

DFI NF4 SLI-D with 414-3 and 310 official bios set to 3.3v psu rail jumper (default)
2x 256MB Gskill PC4400LE does 290-315mhz 2.5-3-3-7 1T on winnie/clawhammer
2x 512MB Gskill PC4400LE does 290mhz 2.5-3-3-7 1T on winnie
2x eVGA 6800U PCI-E SLI 71.89WHL
120GBW D 1200JB
650W Silverstone SLI PSU

6.39 chipset driver
WinXP PRO SP2

did fresh windows install with 6.53 remix drivers and WinXP PRO SP2 and exact same poor ocs

i'm using MCW6002 waterblock with MCP650+MCP600 in series and Thermochill 120.3 with 3x Panaflo 103cfm fans on high speed.

at stock 12x200mhz windows smartguardian idles cpu at 29-31C and load 39-41C

tried using 166 and 180 dividers as well as same problems unstable superpi 1m and pifast errors immediately.

Any ideas ?

[stealth17]

alright well ive done some reading and as i got my 3700 SD today i want to clear some stuff up.

im getting phase-change so should i still remove the IHS?

whats the max safe voltage benching and the max safe 24/7?

has anyone run one on a stock hsf, if so, what did you get?

thanks guys..oh and at say 3.3ghz, does anyone know about how many watts of heat this babby puts out?

[jikdoc]

i've got the same impressions on fx55 vs 3700SD in terms of mem bandwidth. overall i think SD will oc higher than fx55 on single phase. i'm not sure the lower mem bandwidth is real or simply a glitch in the programs that measure it. for example, if you take a look at the 32m scores i posted in the following thread, at nearly the same clocks, the SD is faster even though it's 8mhz slower on cpu and 4mhz slower on ram speed

http://xtremesystems.org/forums/show...598#post856598

@eva, my only guess would be the psu, how is the silverstone? i'm assuming opb is on pcpc 850. have you tweaked drive strength (i know you probably have) also what are opb's ambient temps versus yours?

edit: max spi 1m stable 3.3g (for now, need to sleep)

[joe2004]

My impression is now that SD is not worth one penny over Venice and certainly not $200 more than Venice 3000. As a matter of fact whoever intends to use air only cooling should get Venice.

I can clearly see how SD gets hotter and then drops in performance. I compare the SuperPi iterations on Venice at 2.8 GHz and SD at 2.75 GHz. SD goes faster in the first 5-8 iterations and then starts dropping down to Venice, at the end my Venice chip gets better than SD.
I am sure this is a consequence of heating up.

Now, if I had a phase change this would probably be a different story. Thats an idea.

[snic320]

So i´m on the way of requesting a Sandy 3700+.....you really think it´s not a good deal against the Venice in terms of oc? (will use watercooling and i really dont need to go past 2800Mhzs or so, btw it´s a G4+Bix3+Bix), will mix it with DFI Ultra D and UTT mushkin BLue mems

thanx

[jikdoc]

- 3700+ CABGE 0515SPMW
Is 1.8v safe? Should the IHS come off?

for benching sure
NO!!!

[XTCgr]

I bought OPB's 4000+ Sandy CABHE 0515 SPMW. Anyone experience this ? OPB managed 2850mhz at 1.45v on air using X-90C but i can't get anywhere near that on water ?

superpi 1m stable only with 12x cpu multiplier at:

2700mhz at 1.52v
2760mhz at 1.57v
2808mhz at 1.65-1.67v

my SD 3700+ CABGE 0512 XPMW has the same behaviour (though in my case i can use 11 and 10 multi) - cpu on water and now i am struggling for 2900. I havent tried with air cooling to compare and find out any differences with watercooling